An autonomous driving vehicle uses a global positioning system (GPS) to locate its current position and plans a driving path in accordance with an environmental recognition result based on information obtained from a laser sensor, a vision sensor, and the like.
The autonomous driving vehicle selects an optimal path among previously-defined candidate paths to plan the path. In this case, a sampling technique, which generates the candidate paths corresponding to a road shape and selects the optimal path among the candidate paths, is mainly used to plan the path in real time.
The sampling technique finds more stable and optimal path as the number of the candidate paths increases, but an amount of computation increases to find the optimal path. That is, an optimization of the path and an efficiency of the computation are in a trade-off relation with each other.
Since a conventional path determining method generates a large number of candidate paths and selects the optimal path among the generated candidate paths, the amount of the computation increases, and the efficiency of the computation is lowered.
In addition, another conventional path determining method does not effectively reduce the number of the candidate paths, and thus there is a problem of not being able to select the optimal path.